Proteins play a central role in the signal transmission in living systems since they are able to recognize specific biomolecules acting as cellular receptors, antibodies or enzymes, being themselves recognized by other proteins in protein/protein interactions, or displaying epitopes suitable for antibody binding. In this context, the specific recognition of a given protein unlocks a range of interesting applications in diagnosis and in targeted therapies. Obviously, this role is already fulfilled by antibodies with unquestionable success. However, the design of synthetic artificial systems able to endorse this role is still challenging with a special interest to overcome limitations of antibodies, in particular their production and their stability. Molecular Imprinted Polymers (MIPs) are attractive recognition systems which could be an alternative for the specific capture of proteins in complex biological fluids. MIPs can be considered as biomimetic receptors or antibody mimics displaying artificial paratopes. However, MIPs of proteins remains a challenge due to their large size and conformational flexibility, their complex chemical nature with multiple recognition sites and their low solubility in most organic solvents. Classical MIP synthesis conditions result in large polymeric cavities and unspecific binding sites on the surface. In this review, the potential of the sol-gel process as inorganic polymerization strategy to overcome the drawbacks of protein imprinting is highlighted. Thanks to the mild and biocompatible experimental conditions required and the use of water as a solvent, the inorganic polymerization approach better suited to proteins than organic polymerization. Through numerous examples and applications of MIPs, we proposed a critical evaluation of the parameters that must be carefully controlled to achieve sol-gel protein imprinting (SGPI), including the choice of the monomers taking part in the polymerization.